Where do political prediction polls get their figures?
News reports cite very specific prediction polls about the election with narrow margins of error. How do they conduct these polls and collect accurate information? The only election questions I ever get are from political organizations trying to promote their candidate, never impartial.
Answers ( 1 )
I worked at a research firm that regularly conducted surveys. I don't know the specifics of political polling, but they buy a randomized sample (how truly random it is may differ) of contacts -- mail, email, land line or cell phone. Cell phone is more problematic because it doesn't necessarily correlate to where people live, but voting is. They say it is "likely voters," but I am not sure how that is defined. It may be by their voting history or it may be by respondents saying they are likely to vote. If is is a land line poll of people with a history of voting, it will leave out the majority of youth who are voting for the first time.
IMO, it is important to read the fine print (for instance, on Five Thirty-Eight), and take the polling with a large grain of salt. First, the sample sizes are often small. I have seen national polls of ~1,500 people, which is not large enough to be representative. Second, the respondents are self-selecting. They have the time and interest in answering, so that may naturally rule out groups of busy people who will nevertheless vote. Third, try to find the "raw data" of what the actual question was. There is nearly always interpretation between what they were asked and the conclusion drawn and reported, unless it is a ranking question. Fourth, some people lie. And finally, with the news cycle advancing at such a dizzying speed, many of the polls are outdated in just a few weeks, as some opinions are still shifting with current events.
I think polls are interesting, and sometimes motivating, but they clearly can get the picture wrong a lot of the time, so you have to approach them with a healthy level of skepticism.